What Is a Business?
What is a business, and when does activity for a business begin? Corporate giants like Wal-Mart, General Motors, and Microsoft are unambiguously ongoing and active businesses by any definition. For other cases, it is less clear: Is a trader on eBay a business? Is a software engineer who is currently employed by a software company, but who is considering starting up a spin-off and has engaged actively in some planning, a business? When does a start-up effort cross the threshold between gestation and becoming an operating business? U.S. statistical agencies typically define a business as an entity that is active in terms of having either positive payroll or positive gross revenue, but many other possibilities exist.1 As will become clear, even seemingly straightforward criteria involve a variety of issues associated with the measurement and interpretation of business activity and business units.2
Realizing that businesses are the major source of jobs in the economy,
For example, the Current Population Survey program states that “a business exists when one or more of the following conditions is met: (1) machinery or equipment of substantial value is used in conducting the business; (2) an office, store, or other place of business is maintained, or (3) the business is advertised by listing in the classified section of the telephone book, displaying a sign, or distributing cards or leaflets or otherwise publicizing that the work or service is offered to the general public” (http://www.bls.census.gov/cps/intmanb1.htm).
For discussion of the measurement issues, useful background reading includes Jarmin and Miranda (2002), Spletzer et al. (2004), and Clayton and Spletzer (2005). For discussion of the conceptual issues, see Dunne, Roberts, and Samuelson (1988) and Davis et al. (2006).
as well as of market goods and services, it is important to be able to identify them in a way that is useful for various research and policy purposes. For some applications—such as in measuring national-level output or employment, industry-specific outputs (like radiators for new cars), or the wide range of goods and services provided by a major retailer (like a single megastore providing food, hardware, electronics, and pharmaceuticals within the same structure)—it may be necessary only to identify activities at the firm level, even if that firm operates establishments in several locations. In many other cases, when analysis is focused on understanding the sources of goods, services, or job creation and loss for a given geographic location, it is important to have data on business activities at the establishment level. Physical locations can be identified at the local or regional level, with attention to growth and change for a given neighborhood, county, metropolitan region, or state, or for the sector in which the unit is classified. Such questions require a focus on business entities that make up a geographically defined economy.
In this chapter, we consider these definitional issues in detail. The first part of the chapter focuses broadly on what a business unit is and what entry or exit of a business represents; the remainder of the chapter asks how such concepts are, or could be, measured in the current U.S. statistical system. In that system, there are a number of different units of analysis from which data are collected (see Box 2-1). These include the establishment, the company (firm), the line of business, and the owner. Some of these concepts are related to legal and administrative entities, while others are related to the productive or geographic structure of the firm.3 Given these existing data resources and the possibility of incorporating relatively small modifications, we suggest alternative ideas for collecting data on businesses for the purpose of measuring producer dynamics.
DEFINING BUSINESS UNITS AND IDENTIFYING BIRTHS AND DEATHS
Any procedure for identifying and counting businesses or for providing reliable statistics on business activities is more useful if it begins with a precise conceptualization of a business entity. However, as alluded to above, a number of alternative definitional or conceptual approaches are possible, and they vary in terms of relevance to the production of timely, accurate data on new and young businesses.
Key Measurement Units in Business Statistics
Firm—An organization conducting a business under a legal form of ownership. A firm may operate one place of business or more (such as a chain of restaurants) or have no fixed business address at all (such as a firm represented by a self-employed contractors). Firms may take a number of legal forms (and a single firm may operate more than one type of legal entity):
Sole proprietorship—An unincorporated business owned by an individual. Many businesses run by self-employed persons fall into this category. The business may be the only occupation of an individual or the secondary activity of an individual who also works elsewhere.
Partnership—An unincorporated business owned by two or more persons having a shared financial interest in the business.
Corporation—A legally incorporated business as defined by state laws. These organizations may fall into one of several categories (subchapter S, limited liability, etc.).
An important distinction for data collection purposes is that between employer and nonemployer status:
Nonemployer business—One that has no paid employees but (as typically defined) has positive receipts (for the Census Bureau the threshold is $1,000 or more for most industries and $1 or more for the construction industries) and, as such, is subject to federal income taxes.
Employer business—One that has at least one paid employee (importantly, this triggers inclusion in state UI data systems).
Establishment—A single physical location at which (a firm’s) business is conducted or services or industrial operations are performed.
Line of Business—A group of establishments under common ownership producing the same basic output (perhaps defined by an industry code). This delineating concept typically lies between the firm and the establishment levels.
Industry—A product grouping that facilitates analysis of the relationship among products—such as those defined by the North American Industry Classification System (NAICS).
Primary source: http://www.census.gov/econ/www/index.html
For taxation and other government purposes, businesses are defined in legal terms. In this context, the focus is on businesses registered as independent legal entities—such as a sole proprietorship, general or limited partnership, or one of several corporate forms (subchapter S, limited liability, etc.)—with the appropriate government agency, usually a state department
of commerce, as a specific type of entity. Any action that would be pursued by the courts related to tax payments, enforcing environmental regulations, antitrust legislation, compliance with Equal Employment Opportunity Commission guidelines, and the like would require dealing with the legal entities. Hence, development and administration of legal requirements and regulations require a focus on these kinds of definitions; it is not uncommon, however, for a single business, coordinating resources and strategy, to create several legal entities to gain tax advantages, compartmentalize risk exposure, or facilitate development of financial support.
Alternatively, business entities may be defined in a more purely economic sense. In this context, any buyer or seller of goods and services (not acting solely as a consumer) conducting activities that influence prices or quantities exchanged in a market would qualify as a meaningful unit of analysis. Such business entities could consist of the full- or part-time effort of a single person or of an organization of thousands acting in concert. There has been substantial effort devoted to identifying the number of businesses (or producers) in specific markets and to measuring the dynamics of producer-buyer exchanges.
Assessment of producer dynamics depends on defining a relevant market, and such definitions can be difficult to refine. This difficulty surfaces, for example, in antitrust cases, where market demarcation is often vigorously contested. However, if one collects detailed information on both the geographic location of production and the products and services produced at each location, then the analyst has flexibility in defining markets and hence in constructing measures of producer dynamics. This also relates to the point that a one-size-fits-all definition of business births and deaths will not provide adequate information over the full range of applications. The opening of a Home Depot in a city may represent a new business birth (of an established firm) in that location while, at the national level, this would properly be viewed as internal corporate growth. The statistical system should provide flexibility in allowing accurate measures of business dynamics at the market level, whether that market is local or national in scope. It would clearly be a major benefit if data could be organized around both legal and productive entities.
Legal and Production-Oriented Concepts
To be useful, the definition of a business should provide clarity for determining when a “productive organization” exists, as well as for indicating precise events associated with changes in status (e.g., firm birth or death transitions). The existence of a legal entity is relatively straightforward: one exists when it has been listed in an appropriate registry. However, there is not a one-to-one correspondence between the existence of a legal entity and
the existence of a productive entity. Many legal entities never become operating businesses; many businesses operate for some time before they become legal entities—particularly those representing self-employment. Likewise, many firms persist as a legal entity long after all productive activities have been terminated. And, as noted above, productive organizations may be represented by multiple legal categories. For these reasons, legal status alone is a problematic and incomplete indicator of the presence of a productive organization.
A focus on the impact of businesses on markets would emphasize entities as they affect the prices or quantities of goods and services exchanged. From this perspective, the presence of economic transactions is a critical criterion for defining the birth or death of a firm. For example, an individual’s or team’s efforts to initiate a business start-up, but which have not yet involved any economic exchanges, would be considered personal time allocation but not as an action by an economic actor in any market. A strict application of such criteria would indicate the presence of a productive organization only when an external economic exchange takes place, such as an initial payment for supplies, wages paid to an employee, or the receipt of financial support for business purposes. For example, a personal investment of $1,000 from an owner, placed in a firm checking account, might qualify as a market transaction. Such a conception would not require that the “business” receive any income—only that it engaged in economic transactions, such as the purchase of goods or services in anticipation of the sale of a product. This could lead to the categorization of a large number of start-up efforts, in which preliminary small-scale transactions have taken place, as active productive organizations. For some purposes, such as tracking employment relationships, this may be an appropriate definition of a firm birth—even if the salaries and wages were covered by investors and not by revenues generated from the sale of goods or services.
Issues similar to those described above arise when defining a firm “deactivation.” Firm deactivation or firm exit might be defined by any number of measures of activity (or rather, inactivity): zero employment, zero payroll, zero expenditures, or zero revenue are all possibilities. A challenge here is that many firms accurately characterized by these criteria may still have substantial assets and continue to function for some time, at least as a legal entity. Indeed, some firms may systematically switch back and forth between active and dormant status; for example, consider a sole proprietor in the construction industry who periodically takes on contracts and, in turn, employs workers on a temporary basis only as needed.
A core theme, then, is that entry and exit are not clear-cut terms, and the appropriate definition may reasonably depend on the issue in question and the measurement objective. If the focus is on job creation, then being payroll active is important. Alternatively, if the focus is on the distribution
and dynamics of sales or revenues, then being revenue active is important. The implication is that some flexibility is required conceptually, as well as in terms of measurement, for defining business existence, entry, and exit.
There are two main administrative systems in the United States that define basic business units—the tax system and the unemployment systems. The former defines the basic legal forms of ownership: the primary tax filing entities for businesses are corporations, partnerships, S-corporations, and sole proprietorships (farm versus nonfarm). Each tax-filing entity is given a unique Employer Identification Number (EIN), and these reporting entities form the basis of the Business Master File (BMF) maintained by the Internal Revenue Service (IRS).4 This initial assignment of legal entities, as represented by the unique EINs, is fundamental to the collection of data on U.S. businesses. The BMF defines the universe of all nontax-exempt firms in the United States. Not only does the IRS produce statistics through its Statistics of Income program based on the EIN as the unit of analysis, but also the Census Bureau relies on EINs in their data collection programs.
In the U.S. system, a distinction is commonly made between tax-filing entities that are employers and those that are nonemployers. The latter are businesses with positive revenues but with zero employees (e.g., typically the self-employed or sole proprietors). This distinction is particularly important for data collections based on the Unemployment Insurance (UI) system (ES-202 program), as these include only employer firms. The Bureau of Labor Statistics (BLS) oversees the data collected from the state UI systems and organizes a national database of employers based on them. In the UI system, the legal entity is the UI account. However, through the Multiple Worksite Reports (MWR) system, data on each physical location of employment is obtained from UI account holders, and this allows establishment-level data to be constructed.
While the legal concept of a business as defined by the IRS is central to the U.S. statistical system, the EIN as a unit of analysis or as a statistical sampling unit is often inadequate. As noted above, the EIN is an administrative entity that may or may not correspond to a common conception of an ongoing business as a unified, centrally coordinated productive organization. First, the existence of an EIN does not necessarily mean the entity is an ongoing business. For example, an entity with the potential to become a
firm may apply for an EIN with the plan of starting a business, but that business may never actually come into being. The EIN would exist but no taxable business activity would be recorded. Similarly, a business that ceases to exist may still retain an EIN but, again, no business activity may be present in the firm. These situations are readily handled by requiring a certain level of revenue, payroll, or employment to classify as an ongoing business. That said, the choice of threshold variable and the unit of time for which the variable is being measured (e.g., monthly employment or annual payroll) can affect whether an entity is viewed as active or inactive. Second, the EINs for a tax entity can change for both economic reasons and for administrative reasons. A company’s EIN(s) may change due to acquisition—the company comes under the legal control of another entity—or due to some administrative tax action such as a change in the legal form of the organization. Third, and most important, the EIN-level unit of analysis may not be appropriate for measuring the economic concepts of interest. This is particularly true in the case of complex business organizations whose EIN-based data may span a range of geographic and product markets.5
In order to produce detailed statistics of economic activity by geographic location and by industry, company-based data represented by the EIN unit must be disaggregated into finer units of analysis that make both economic and accounting sense. Two units of analysis have typically been used to measure economic activity below the company level—the establishment concept and the line-of-business concept.
The establishment is the primary unit of analysis underlying business data collected and maintained by the U.S. statistical agencies.6 The Census Bureau defines an establishment to be “a single physical location where business is conducted or where services or industrial operations are performed.” In the employer universe (which is the basis for many of the surveys produced by the Census Bureau), there is an additional requirement that the establishment must have one or more employees and be in operation during some part of the year. Firms without employees are referred to
as “nonemployers” and are often treated separately in Census Bureau data collection programs.
The Census Bureau also allows a physical location to be disaggregated into multiple establishments if there are distinct lines of operation producing different goods (typically goods that are in different industries) at the same location. This disaggregation of a physical location is done only for cases in which the activity in at least two production lines falls into different industry sectors and output can be allocated accordingly. This typically involves large vertically integrated production locations. However, this disaggregation of physical locations complicates the picture because a small, integrated physical entity may be treated simply as one establishment while a large integrated one (doing essentially the same things) may be treated as multiple establishments.
BLS employs a similar establishment concept in the Quarterly Census of Employment and Wages (QCEW) program used to support its business register. Although there are some differences in industry coverage across the Census Bureau and BLS programs, the notion of an establishment is quite similar. One definitional difference that does exist between the agencies has to do with the treatment of very small multiunit operations. While the Census Bureau makes no distinction for reporting based on the size of small multiunit establishments, the QCEW allows small multilocation employers (10 or fewer employees in secondary work sites within the state) to file a combined report as a single establishment.
An alternative unit of analysis, used less frequently than the establishment concept, is the line of business. The line of business typically lies between the company level and the establishment and is organized around production in a sector or industry. The Federal Trade Commission line-of-business data program (discontinued) required companies to disaggregate their financial data in this way. In some Census Bureau surveys, large firms are asked to report for a particular line of business. A statistical reporting unit based on line of business may represent the activity (and even partial activity) of a number of individual establishments. Alternatively, a firm may be directly surveyed about economic activity for an industry or collection of industries that, again, may span a set of individual establishments owned by the firm. For example, the Annual Capital Expenditures Survey of the Census Bureau surveys firms regarding capital expenditures, but it requires firms to break out the expenditures by industry.
DEFINING BUSINESS UNITS FOR THE PURPOSE OF MEASURING DYNAMICS
The key challenges for a data system intended to have the capacity to measure producer dynamics involve accurately tracking (1) the entry,
growth, and exit of business units and (2) the employment flows generated by these expansions and contractions and births and deaths of employers (Davis, Haltiwanger, and Schuh, 1996). The business dynamics literature generally has measured job creation and job destruction by tracking changes in employment in individual establishments over time. Data series on job creation and destruction have been constructed by industry, region, and characteristics of producers, such as size and age. One outgrowth of this line of research in the United States has been the development of the Business Employment Dynamics program at BLS (Spletzer et al., 2004) and the Quarterly Workforce Indicators (QWI), a data series published since 2003 by the Census Bureau that offers detailed information on local labor market dynamics. Each of these programs provides timely information using QCEW data on job creation in new and growing establishments and job destruction in exiting and shrinking establishments. The QWI data series from the Census Bureau also provides measures of worker hires and separations classified by worker characteristics, such as gender and age.
A second approach to defining business units takes an industrial organization perspective and focuses on measuring producer participation in a given market or industry. In this approach, participation in a specific product market by a firm is key to defining the business unit. The specific market may be defined along a range of dimensions including, but not necessarily limited to, industry and geography. The market may correspond to an industry or product that is national (or international) in scope (e.g., semiconductors), or it may be much more localized, as would be the case for, say, the services of most restaurants. Under this definition, a firm or an establishment may produce for a single market or for multiple markets, and it is fundamental to be able to measure a firm’s participation in each one.
The business unit described in a market-oriented approach may, in some cases, represent a nondiversified firm that operates in a single market. Here, the legal definition of the firm corresponds to our notion of the business unit, and the standard administrative or tax entity is an adequate statistical unit. Examples include a manufacturing firm operating at a single location producing a specialized good, or a service provider (like a dentist) that sells in a local market. In both cases, these are single establishment firms selling in a single market. However, many firms produce in multiple industries and operate across many distinct geographic markets. The link, in these cases, between the business unit and the firm is not one-to-one, and thus the firm definition is inadequate for measuring participation in markets. It is important to recognize that, while such multimarket firms may be relatively small in number compared with single-market firms, they typically represent a substantial fraction of economic activity because of their large sizes.
From a cross-sectional perspective, there are a number of key charac-
teristics needed to identify a business unit operating in a product market. For one, the ownership structure of the assets owned by a firm must be transparent in the data. That is, the ownership links between the parent company and owned assets, such as establishments, subsidiaries, and lines of business, must be clearly identified. This is necessary in order to determine the number of distinct decision makers in the market. For example, a firm may operate two individual establishments that produce for the same market. Under the product market definition proposed above, one would not want to count such establishments as two independent business units if they are under common ownership. Therefore, each business unit must be identified with the parent firm that owns the unit.
In addition to information on ownership structure, it is important to be able to identify the products and services produced by each firm. Traditionally, this need has been addressed by focusing on identifying industries of operation, enumerating products produced, defining classes of customers, or providing information on the selling format. Across sectors of the economy, different information is required in order to classify the economic activity of producers, as well as to identify the market they sell in. The information required from a retailer is different from that required from a manufacturer. For many sectors of the economy, the location of production is central to defining markets and thus business units, as well. Retail, service, construction, and even manufacturing firms often sell in local markets. For firms in national markets, knowing where they produce may not be critical to defining the business unit; however, it may be important for understanding the impact of the business unit on local input markets (e.g., labor markets). Finally, along with identifying the activities that a producer undertakes, a measure of the economic importance of the activity is required. Shipments, revenues, or sales volumes are likely candidates, although such information is less readily available at the establishment level than it is for employment and payroll.
The ability to measure the entry and exit of market participants, along with changes in the performance of incumbent producers, is fundamental to understanding how markets evolve. Again, our definitions are centered on participation in a product market. An entrant is defined generally as a firm that is producing in a market in the current period (quarter or year) that was not producing in the market in the prior period. An exit is similarly defined as a firm that was producing in a market in the prior period but that is not producing in the market in the current period. For many purposes, one would ideally like as high-frequency data on market entry and exit as possible, perhaps quarterly or, at a minimum, annually. However, it must
also be recognized that attempts to generate data at a higher frequency for entering firms may come with a cost. Identifying the industry, geographic location, ownership, and other detailed characteristics of a business often requires surveys and therefore takes time; at the time of entry, less detailed or accurate information may be all that is available.
On the entry side, it is important to consider both de novo and existing firms. A de novo entrant is a new firm producing in a market and is what comes to mind when thinking about entrepreneurial processes. In such cases, the firm did not produce in any other markets prior to its entry into the market of interest. Many single-location births or new nonemployer firms fit into this category. However, another important source of entry into markets is existing firms diversifying into new markets (Dunne, Roberts, and Samuelson, 1988). These existing firms may enter a market by shifting existing assets from one industry to another or by creating a new production facility to service the market. In these cases, the firm may be established, but it is a new participant in a specific market or sector. What is fundamental here is that, in order to fully measure entry, one must be able to identify not only de novo entrants (new firms created for this market), but also the expansion of existing firms participating in a market new to that firm. Similarly, the exit of a producer from a market may correspond to the death of a firm, or it may simply represent a shift in the mix of production occurring at an ongoing firm. Under this definition, the closure of a plant or store may or may not represent exit from a market. These differences in the types of exit also should be distinguishable in data used to study producer dynamics.
In order to measure de novo entrants, one needs reliable and timely information on new firm formation. This information typically comes from administrative sources, such as those underpinning the tax and UI systems. Identifying when existing producers diversify into new markets is in one way more complicated, as it requires longitudinal information on the distribution of economic activity (e.g., along industry or geographic dimensions) of incumbent producers. These definitions of entry and exit require that the ownership of business units is tracked accurately, that industry and product coding is uniform, that the output is measured consistently and, sometimes, that the location of business units is tracked.
Regarding the ownership issues, each business unit must be identified with the firm that owns it at any given point in time, and changes in ownership must be accurately tracked. Ownership and administrative changes that do not affect the business operation but do change the legal entity must not be mistakenly picked up as an exit of an existing firm and the entry of a new firm. Vale (2006) found that an important difference affecting the international comparability of business start-up statistics internationally relates to how countries treat administrative and ownership
changes in the data. Reactivations and reregistrations of businesses can result in substantial overestimation of business start-ups. Baldwin, Beckstead, and Girard (2002) report that business start-up rates are substantially overestimated, both in numbers and employment shares, when one relies simply on tracking firms in a business register. Many new startups are, in fact, unlinked acquisitions. However, ownership changes that affect the number of competitors in a market need to be reflected in the data. A horizontal merger between two competitors results in a reduction in the number of independent business units in the market and thus represents a form of exit (but not deactivation of a production unit).
Detailed information on business location may also be vital, depending on the industry or issue under study. Clearly, in such industries as retail and construction—and sometimes even in manufacturing—products markets are local in nature. In order to characterize the entry and exit of firms into these typically narrow geographic markets, detailed information on the location of operations is required. Alternatively, for some industries, such as consulting and perhaps information technology services, the location of the business unit may be effectively undefined. On balance, however, it is probably useful to think of business location as an important characteristic for helping to link business units to specific markets. Many applications require that longitudinal data on the locations of a firm’s business activities correspond to the existing definition of an establishment used in current business registers.
Given the market-oriented emphasis in our conceptual definitions of business units, it is clear that high-quality, consistent industry and product coding are also required. While measures of entry and exit of producers in a market should not be driven by changes in industry coding systems, actual shifts by a firm out of some industries and into others should be reflected in data on the creation and destruction of business units (e.g., lines of businesses) within the firm. This requires, among other things, improving and updating product codes to reflect an increasing share of economic activity in dynamic service areas and high-tech manufacturing. This kind of accuracy can be difficult to achieve for new and very small firms because they are constantly changing; thus, one cannot expect to immediately have fully up-to-date coding for some fraction of the business population.
With respect to measuring the dynamics of ongoing business units, precise, standardized records of the same businesses over time are clearly a major asset. As discussed above, inability to accurately track incumbent businesses invariably leads to measurement errors in business birth and death statistics. For these ongoing entities, some measure of size in the market—such as sales, revenue, or output volume—is required so that the growth of incumbent firms can be tracked.
Beyond the period-to-period transitions of firms, analysts of producer
dynamics are often concerned with tracking the history of producers. For example, it is well established that the growth process for young producers differs from that of older producers. However, this process is quite different for new business units that are owned by older firms compared with de novo firms created to enter the market. New businesses owned by older firms are typically larger and have less volatile growth rates and lower failure rates than de novo entrants. In order to develop rich histories and summary statistics on firm dynamics, information on both the age of a business unit and the age of a firm that owns the business unit is useful. Age should typically be defined as an establishment age, and this should be tracked through changes in ownership. Ownership or administrative changes should not generate a new establishment age.
For some purposes, firm age is important but more difficult to measure. Consider, for example, the purchase of an old establishment by a younger firm, in which case the latter becomes the owner if it is the ongoing legal entity. Alternatively, when an older firm purchases a young establishment, the latter should be assigned the age of the purchasing firm age after the transaction. This is similar to a new establishment being opened by an old firm. The establishment age is set to when it opens, but the firm age is based on when the firm opened its first establishment. The key point is that establishment age is clear-cut, defined by the initial period in which it appears in the sample frame. Firm age requires a more detailed enumeration of cases, and rules can be established for this purpose.
Identifying Nascent Businesses
For the purpose of producing many economic statistics, the focus of business data collection must necessarily be on fully operating entities. That said, business creation processes are of great interest, and the conceptual framework for constructing data on these entities is far less developed. In thinking about firm creation, it is useful to begin by defining two key transitions: that which occurs when one or more persons begin to mobilize time and resources to implement a new firm—the beginning of the gestation or firm creation process—and that which occurs when the start-up effort can be considered a going business. There are, at present, no ongoing federal data collection programs that focus on tracking either transition in a representative sample of household or business populations. The Current Population Survey has been used to identify individuals reporting a sudden increase in time devoted to self-employment activities, an indicator of individual business creation activity. But the operational definition—identifying an increase in effort from two consecutive monthly reports of 15 hours per week on new self-employment initiatives—has an unknown relationship to the actual provision of goods and services, job creation, or the
implementation of a going business (Fairlie, 2006). It may be that modest adjustments to the Current Population Survey or other ongoing household-based surveys, such as the American Community Survey program, would provide a basis for estimating the participation of U.S. adults in activities associated with the creation of new businesses.7
An ongoing research program—the Global Entrepreneurship Monitor (detailed in Appendix A), which collects data on various aspects of entrepreneurship through a series of coordinated household surveys in a number of countries—has produced operational definitions of start-up transitions. Representative samples of adults have been used to locate individuals that appear to have initiated the creation of a new firm, either on their own or for their employer. Three criteria are used to identify nascent entrepreneurs: (1) they consider themselves involved in new firm creation, (2) they report being actively engaged in behavior to create a new firm, and (3) they expect to own part of the new firm.8 To separate those in the start-up process from those owning and managing new firms, two criteria have been developed to represent the “firm birth” transition. The operationally more complex one is to identify those that report positive monthly cash flow covering all expenses and salaries for more than three months; a slightly simpler version identifies those reporting salaries and wages paid to the owners for more than three months. There is some correspondence between these criteria and those commonly used to identify new entrants in business registers maintained by the IRS, BLS, Census Bureau, and Dun & Bradstreet. However, inclusion in these registers takes place at different stages of the firm creation process, reflecting considerable variation in the sequence of major start-up events—for example, significant financial investments, hiring the first employees, expenditure on capital and operations, initial sales, and initial profits (Reynolds, 2007).
It has been suggested that, for venture capital-sponsored firms, revenue generation happens, on average, later in a firm’s start-up phase than employment creation (Kaplan, Sensoy, and Strömborg, 2005). In such cases,
Substantial work on effective, cost-efficient procedures for identifying nascent entrepreneurs in household surveys has been completed as part of the University of Michigan administered the Panel Study of Entrepreneurial Dynamics (PSED) research program (http://www.psed.isr.umich.edu). Screening completed to identify the PSED II cohort in the fall of 2005 involved adding a 2-minute module to 34,000 household interviews to locate 1,200 active nascent entrepreneurs (a yield of about 35 nascent entrepreneurs per 1,000 households, and about 2,000 minutes of interview time).
These criteria have been employed in dozens of samples in the United States and in over 40 different countries in all stages of development. For U.S. applications, see the appendices in Gartner, Carter, and Reynolds (2004); for cross-national procedures, refer to Reynolds et al. (2005).
employment-based data on existence would measure entry earlier than would the product market participation approach discussed above. Others have emphasized the fact that revenue generation by small firms often precedes the hiring of the first employee and, in fact, many firms never hire an employee through their entire lifetimes (Davis et al., 2006).
While the product market approach leads to a focus on firms from the birth event forward, it has become clear that there is a lack of information and attention to the mechanisms and processes that precede it. Indeed, it is now clear that a substantial amount of time (“sweat equity”) as well as resources (mainly informal funding) is absorbed by start-up efforts that never become operating businesses by any criteria. In addition to substantial intellectual interest, a more complete understanding of the start-up process itself has substantial implications for regional economic policies as well as personal career planning.
CONCEPT VERSUS EXISTING DATA COLLECTION
In the current statistical system, the unit of analysis that maps most flexibly into measurement needs is based on establishment reporting. Under ideal conditions, information on the range of products or services produced at each business location is also provided so that economic activity can be classified into the relevant market(s). The data collected in the quinquennial economic censuses by the Census Bureau have this product- and business-type detail for surveyed establishments.9 In addition, the data collection procedures of both BLS and the Census Bureau allow establishments to be disaggregated into distinct reporting units if they are important producers of multiple products.
A weak point embedded in both agency procedures has to do with the collection of data from small establishments owned by multiunit firms. The annual Company Organization Survey (described in Appendix A) omits small multiunit firms; thus the Census Bureau captures only the entry and exit of such smaller multiunit establishments owned by small multiunit firms in the quinquennial economic censuses. The MWR do not require an employer to disaggregate the data by establishment until 10 or more employees work away from the parent location. In each case, when dealing with small multiestablishment firms, it may be difficult to measure producer dynamics at the establishment level, especially at high frequencies.
A key requirement of the ideal business unit definition proposed above is that a measure of output be collected by each business in each period. In practice, this would be expensive, and it is not clear that it would be practical. Certainly, greater use of IRS data on sales and less reliance on administrative data in the economic census would be helpful on this score. Ideals aside, it still must be pointed out that, within the current U.S. statistical and administrative data systems, output data are collected in a very heterogeneous fashion across sectors, across time, and across different types of producers. Every five years, through the economic census programs of the Census Bureau, data on revenue and sales are collected for a wide range of sectors at the establishment level. At an annual frequency, data on revenues are collected by tax authorities on an EIN basis, but these data are not available to all the statistical agencies.10 Moreover, for larger organizations, revenues at the EIN level are not particularly useful for measuring producer dynamics at the market level.
To be sure, the Census Bureau has a large number of surveys that collect information on output, revenue, and sales that are more timely than the economic census data. Surveys in manufacturing, services, and retail cover large segments of the economy and provide data at annual and subannual frequencies (quarterly and monthly). However, this coverage is less comprehensive, particularly at the higher frequencies, and the surveys focus on providing accurate estimates of sales or output at a point in time at the aggregate level. The design of these surveys focuses on larger units and is not geared to produce dynamics statistics, which require high-quality information on smaller firms and accurate tracking of individual business units over time.11
The administrative data generated through the UI system measure employment and payroll systematically for a large number of sectors, at a high frequency (quarterly), and at the establishment level, but they do not contain information on output. These data facilitate production of high-frequency information on producer dynamics at the establishment level
using employment or payroll data. In these data, establishments are classified by industry, but there is no information available on product detail. Good geographic detail is provided down at least to the county level (and is increasingly available by latitude and longitude). However, defining market participation with these data is somewhat more difficult, since ownership links across establishments are not comprehensive. Under the definition of market entry proposed above, the opening up of a new operation by an existing producer may or may not constitute entry, depending on whether the existing producer is already an incumbent in the market. Because ownership links are not comprehensive in the UI data (especially across states), it may not be possible to classify entry and exit by existing firms in the manner proposed above for certain industries. However, these data do provide the ability to generate establishment-level statistics on entry, exit, and incumbent growth.
A promising new data source that allows for the measurement of producer dynamics is being developed by the Longitudinal Employer-Household Dynamics (LEHD) program at the Census Bureau. These data have some distinct advantages over the existing data sources with respect to measuring and tracking business unit dynamics. LEHD researchers have linked employer data with employee data from a set of states in the UI system. The capability to track both employers and employees allows for more detailed data to be constructed on job creation and destruction. It also allows for the development of measurement algorithms to improve identification of the entry and exit of business units and to relate these new or closing business units to continuing business units by tracking the flow of workers across them.12 These data have the potential to better describe the entry and exit process of multiindustry and multilocation firms.13 Moreover, by linking these data to the Census Bureau business register, comprehensive ownership links across establishments can often be constructed.14 This will generally allow data users to distinguish between de novo and experienced firm entry and exit.
This discussion of the measurement of business activity and entry and exit has focused on the measurement of core concepts like revenue, payroll, employment, product mix, and location. There are other measures of business activity that are important in their own right, as well as for quantifying business entry and exit. One example is physical capital and expenditures. A new business entity may have capital expenditures and assets prior to the onset of revenue, payroll, and employment. As such, the measures of capital expenditures or assets could be used, by some definition, to identify business entry. In addition, measurement of the means by which business activities (for capital expenditures or other expenditures) are financed is extremely important for understanding the processes underlying business formation and development.
In this chapter, we discuss a number of different definitions of businesses and how the definition must be appropriate for the use to which it is being applied. If a business frame is maintained with detailed information on the geographic location of production, the products and services produced at each location, the ownership structure of the firm, and the ability to track production units over time, then the measures of producer dynamics outlined above can be constructed. The unit of observation that comes closest to providing the necessary features—and the one that often is and should be used in most statistical agency programs—is the establishment definition. Establishments are physical locations in which economic activity is assigned industry code(s) and whose ownership relationships can be identified. Establishment data with features such as those described above provide the overall flexibility required to measure producer dynamics across a wide range of firm types, from small start-up firms to large expanding and contracting ones. The opening of a new firm and the opening of a new branch of an existing firm can also be distinguished with such data.
The establishment definition also allows flexibility in measuring producer dynamics along geographic dimensions. For some applications, local data on entry, exit, and growth are required. Establishment-level data allow for the construction of such statistics, whereas company or line-of-business data generally do not. Detailed industry coding of establishments would enhance the types of statistics that can be produced. Ideally, it would be useful, for certain industries and types of producers (large firms), to have information on the range of industries (products) in which an establishment operates. Finally, there is justification for using input measures (i.e., employment or payroll) or output measures (i.e., revenues or shipments) as the basic measure to judge economic activity and to create measurement thresholds for defining an active business. An output-based litmus, such as rev-
enues, has an advantage in that it is a basic measure of both employer and nonemployer business activity and is a likely choice to measure aspects of producer dynamics for all segments of the business population.
The bottom line is that establishment data are integral to a flexible business data system, and the statistical agencies should resist giving this up simply because survey respondents do not necessarily organize their data in this particular format. Furthermore, calling mainly for use of the establishment concept in defining a business is a practical statement that aligns with what the statistical agencies can reasonably collect: data on location, industries of operation, ownership links, and economic activity. The framework is already used to generate statistics on producer dynamics; its full development requires that what is already measured (surveyed) in economic censuses should be extended to a larger set of firms in the business population on an occasional basis.